The Meta team has entered AI CICERO into an anonymous online diplomacy gaming league. The AI played 40 games, including an 8-game tournament with 21 players; CICERO placed first in the tournament. For all 40 games, CICERO ranked in the top 10% of players with an average score of 25.8%, while the average score of his 82 human opponents was 12.4%.

What is AI CICERO?

Meta AI Research recently provided CICERO, an artificial intelligence that can beat most humans at the strategy game Diplomacy, a game that requires coordinating plans with other players. CICERO combines chatbot-like dialogue capabilities with strategic reasoning and recently placed first in an online Diplomacy tournament against human players.

Games have long been a testing ground for new AI breakthroughs, from Deep Blue’s victory over chess grandmaster Garry Kasparov, to AlphaGo’s mastery of the game of Go, to Pluribus beating the best humans in poker. Truly useful and versatile players will need to go beyond simply moving pieces around the board. Can we create more effective and flexible agents that can use language to negotiate, persuade, and work with people to achieve strategic goals, similar to the way humans do?

Unlike games like Chess and Go, diplomacy is a game about people, not pieces. If a player can’t recognize that someone is probably bluffing or that another player would perceive a particular move as aggressive, they will quickly lose the game. Likewise, if he doesn’t talk like a real person – showing empathy, building relationships, and speaking with knowledge of the game – he won’t find other players willing to work with him.

Key to the achievement is the development of new techniques at the intersection of two completely different areas of AI research: strategic thinking, used in players like AlphaGo and Pluribus, and natural language processing, used in models like GPT-3, BlenderBot 3, LaMDA, and OPT-175B. CICERO can, for example, infer that it will need the support of one particular player later in the game, and then develop a strategy to gain that person’s favor – and even recognize the risks and opportunities that player sees from its particular perspective.

Tags: , , , , , , , , , , , , , , , , , , , ,
Editor @ DevStyleR